On Fri, Apr 13, 2018 at 7:36 AM, Commit Bot <b...@cloudius-systems.com>
wrote:
> From: Wang Yu <yuw...@linux.alibaba.com>
> Committer: Waldemar Kozaczuk <jwkozac...@gmail.com>
> Branch: master
>
> add boot_time event of uncompress lzloader.elf
>
> sometimes i found uncompres cost serival times,
> so add boot_time event of uncompress lzloader.elf
>
Excellent. The important benefit of this patch is that it stops
blaming the "TLS initialization" for the boot time that is really
taken up by the uncompression (I wonder if it's all uncompression?
but it's definitely not TLS initialization, which as you show only
takes a fraction of a millisecond).
> after patch
> ...
> disk read (real mode): 45.09ms, (+45.09ms)
> uncompress lzloader.elf: 76.90ms, (+31.81ms)
> TLS initialization: 77.65ms, (+0.75ms)
> .init functions: 82.69ms, (+5.04ms)
> SMP launched: 83.72ms, (+1.03ms)
> VFS initialized: 88.67ms, (+4.95ms)
> ...
>
> Signed-off-by: Wang Yu <yuw...@linux.alibaba.com>
> Message-Id: <1523516894-14812-2-git-send-email-yuw...@linux.alibaba.com>
>
> ---
> diff --git a/arch/x64/arch-setup.cc b/arch/x64/arch-setup.cc
> --- a/arch/x64/arch-setup.cc
> +++ b/arch/x64/arch-setup.cc
> @@ -50,6 +50,7 @@ struct osv_multiboot_info_type {
> struct multiboot_info_type mb;
> u32 tsc_init, tsc_init_hi;
> u32 tsc_disk_done, tsc_disk_done_hi;
> + u32 tsc_uncompress_done, tsc_uncompress_done_hi;
> u8 disk_err;
> } __attribute__((packed));
>
> @@ -140,6 +141,10 @@ void arch_setup_free_memory()
> time = (time << 32) | omb.tsc_disk_done;
> boot_time.event(1, "disk read (real mode)", time );
>
> + time = omb.tsc_uncompress_done_hi;
> + time = (time << 32) | omb.tsc_uncompress_done;
> + boot_time.event(2, "uncompress lzloader.elf", time );
> +
> auto c = processor::cpuid(0x80000000);
> if (c.a >= 0x80000008) {
> c = processor::cpuid(0x80000008);
> diff --git a/arch/x64/boot16.S b/arch/x64/boot16.S
> --- a/arch/x64/boot16.S
> +++ b/arch/x64/boot16.S
> @@ -20,7 +20,9 @@ mb_tsc1_lo = (mb_info + 88)
> mb_tsc1_hi = (mb_info + 88 + 4)
> mb_tsc_disk_lo = (mb_info + 88 + 8)
> mb_tsc_disk_hi = (mb_info + 88 + 12)
> -mb_disk_err = (mb_info + 88 + 16)
> +mb_uncompress_lo = (mb_info + 88 + 16)
> +mb_uncompress_hi = (mb_info + 88 + 20)
> +mb_disk_err = (mb_info + 88 + 24)
> mb_cmdline = (mb_info + 16)
> mb_mmap_len = (mb_info + 44)
> mb_mmap_addr = (mb_info + 48)
> @@ -161,6 +163,9 @@ done_e820:
> mov %eax, %fs
> mov %eax, %ss
> call *lzentry
> + rdtsc
> + mov %eax, mb_uncompress_lo
> + mov %edx, mb_uncompress_hi
> mov $loader, %eax
> mov $mb_info, %ebx
> call *entry
> diff --git a/include/osv/boot.hh b/include/osv/boot.hh
> --- a/include/osv/boot.hh
> +++ b/include/osv/boot.hh
> @@ -21,7 +21,7 @@ private:
> // relatively late (the code that takes the measure is so early it
> cannot
> // call this one directly. Therefore, the measurements would appear
> in the
> // middle of the list, and we want to preserve order.
> - int _event = 3;
> + int _event = 4;
> time_element arrays[16];
>
> void print_one_time(int index);
> diff --git a/loader.cc b/loader.cc
> --- a/loader.cc
> +++ b/loader.cc
> @@ -109,7 +109,7 @@ void premain()
> }
>
> setup_tls(inittab);
> - boot_time.event(2,"TLS initialization");
> + boot_time.event(3,"TLS initialization");
> for (auto init = inittab.start; init < inittab.start + inittab.count;
> ++init) {
> (*init)();
> }
>
> --
> You received this message because you are subscribed to the Google Groups
> "OSv Development" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to osv-dev+unsubscr...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.
>
--
You received this message because you are subscribed to the Google Groups "OSv
Development" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to osv-dev+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
